sandeepmistry
Optimize MCP2515Class::endPacket() by up to 2x
Optimize MCP2515Class::endPacket()
-
Use LOAD TX BUFFER instruction to set TXBn* registers sequentially, as well as to write into the TX data registers. This is done with a single CS pull, and requires sending N+6 bytes over SPI. The old code was doing N+5 writeRegister() calls, each doing 1 CS pull and sending 3 bytes.
-
Use the 1-byte RTS SPI instruction instead of WRITEing to TXBnCTRL. This reduces the number of bytes sent over SPI by 2.
For N = 8,
- we now send 15 bytes over SPI before the while(...TXBnCTRL...) loop vs 42 (2.8x reduction)
- we do just 2 CS pulls before the while(...TXBnCTRL...) loop vs 13 (6.5x reduction)
We still do at least 3 CS pulls and send at least 10 bytes over SPI for the condition of the while(...TXBnCTRL...) loop, to clear TXnIF and to check for errors.
That means for N = 8, in the case when we do zero iterations of the while loop, we end up sending 25 bytes over SPI instead of 52, which is >2x reduction.
